Research on Bank Customer Churn Prediction Using Machine Learning

In today’s highly competitive banking sector, customer churn poses a significant challenge, directly affecting profitability and customer retention efforts. This research aims to develop a predictive model for customer churn using advanced machine learning techniques. A comparative analysis of multiple supervised learning algorithms — including Logistic Regression, Decision Tree, Naive Bayes, K-Nearest Neighbors (KNN), XGBoost, and Random Forest — was conducted on a publicly available dataset from Kaggle. Additionally, deep learning techniques using Artificial Neural Networks (ANN) were implemented through TensorFlow and Keras frameworks. The study emphasizes the importance of feature engineering and data preprocessing strategies such as oversampling and undersampling to handle class imbalance. Among all the models evaluated, the Random Forest classifier achieved the highest accuracy of approximately 87%, proving to be the most robust and stable model for churn prediction. The results highlight key factors influencing churn, such as customer age and account activity, providing actionable insights for banks to enhance customer engagement and reduce attrition.